Hermes Agent users eye multimodal desktop

This is the right direction for local AI, but the stack is still split across too many pieces: model, client, vision, and tool integrations all need to line up before it feels like a true daily driver.

• –Hermes Agent already supports multimodal vision, but it still depends on pairing it with a client that handles screenshots and MCP cleanly.
• –Anytype’s MCP server makes the knowledge-base side viable; the gap is a polished desktop UX for non-coding workflows.
• –On this hardware, the practical path is usually a strong text model plus a lighter vision model, not one giant local model doing everything.
• –The real product opportunity is a “life admin” assistant with the friction hidden, not another coding-centric agent UI.